With the growth of cloud services, concerns have been raised regarding illegal sharing of the commercial video. To prevent the illegal sharing automatically, the method for classifying video as `commercial' or `noncommercial' is essentially required. Since most commercial video has a logo as a visible watermark, automatic logo detection can be an efficient method for the video classification. In this paper, we present an improved logo detection method which correctly detects the logo in any types of video using learning-based logo verification. Experimental results show that the proposed method achieves improved detection performance as compared with the existing method, and thus can be effectively used for classifying the video.